Method for detecting white matter lesions based on medical image

Abstract

According to an embodiment of the present disclosure, a method of detecting a white matter lesion based on a medical image performed by a computing device is disclosed. The method may include: receiving a medical image including at least one brain region; and estimating a first white matter lesion and a second white matter lesion based on the medical image using a pre-trained neural network model.

1 . A method for detecting white matter lesions based on medical image, performed by a computing device including one or more processors, comprising:
 receiving a medical image including at least one brain region; and   generating information for estimating a first white matter lesion and a second white matter lesion based on the medical image using a pre-trained neural network model,   wherein the generating the information for estimating the first white matter lesion and the second white matter lesion includes:   extracting a brain region including white matter from the medical image, wherein the brain region is a region from which a bone region of a brain has been removed; and   wherein the information is generated based on the extracted brain region.   
     
     
         2 . The method of  claim 1 , wherein the first white matter lesion is estimated based on periventricular white matter hyperintensity, and the second white matter lesion is estimated based on deep white matter hyperintensity. 
     
     
         3 . The method of  claim 1 , wherein the generating the information for estimating the first white matter lesion and the second white matter lesion further includes:
 generating first information for estimating the first white matter lesion based on the extracted brain region, using a first lesion estimation module included in the neural network model; and   generating second information for estimating the second white matter lesion based on the extracted brain region, using a second lesion estimation module included in the neural network model.   
     
     
         4 . The method of  claim 3 , wherein the first information is generated based on the extracted brain region using the first lesion estimating module, and the second information is generated based on the extracted brain region using the second lesion estimating module. 
     
     
         5 . The method of  claim 1 , wherein the neural network model is trained based on an image in which true label is greater than a selected number of pixels among input images. 
     
     
         6 . The method of  claim 1 , wherein the neural network model is trained using a cross-entropy loss function with smoothed weights based on a log function. 
     
     
         7 . The method of  claim 1 , further comprising:
 generating a first mask including information for the first white matter lesion; and   generating a second mask including information for the second white matter lesion.   
     
     
         8 . The method of  claim 7 , further comprising:
 generating a third mask including information for an entire lesion in which a white matter is degenerated, by matching the first mask and the second mask.   
     
     
         9 . The method of  claim 1 , further comprising:
 generating either first volume information for the first white matter lesion or second volume information for the second white matter lesion, based on information for voxel dimension of the medical image and a number of selected units for each of the estimated lesions.   
     
     
         10 . The method of  claim 1 , further comprising:
 calculating a severity of each of the first white matter lesion or the second white matter lesion, based on an index indicating a severity of a white matter degeneration.   
     
     
         11 . The method of  claim 1 , further comprising:
 generating a user interface including a mask generated based on each of the first white matter lesion and the second white matter lesion, volume information, and degeneration severity information, and   wherein the user interface includes:   a first region for displaying an image of a lesion in which a white matter is degenerated, based on the mask generated based on each of the first white matter lesion and the second white matter lesion; or   a second region for displaying at least one of the volume information and the degeneration severity information.   
     
     
         12 . The method of  claim 11 , wherein each of the first white matter lesion and the second white matter lesion is displayed in different colors, in the first region. 
     
     
         13 . A computing device for detecting white matter lesions based on medical image, comprising:
 a processor including at least one core;   a memory including program codes executable in the processor; and   a network circuit for receiving a medical image including at least one brain region,   wherein the processor is configured to:   generate information for estimating a first white matter lesion and a second white matter lesion based on the medical image using a pre-trained neural network model,   wherein the at least one brain region includes white matter, wherein the at least one brain region is a region from which a bone region of a brain has been removed,   wherein the information is generated based on the extracted brain region.   
     
     
         14 . A user terminal, comprising:
 a processor including at least one core;   a memory;   a network circuit for receiving a user interface including information for a lesion in which a white matter is degenerated, from a computing device; and   an output circuit for providing the user interface,   wherein the user interface comprising at least one of the following:   a first region for displaying an image of the lesion in which the white matter is degenerated, based on the mask generated by the computing device based on each of a first white matter lesion and a second white matter lesion; or   a second region for displaying at least one of volume information and degeneration severity information, which are generated by the computing device based on each of the first white matter lesion and the second white matter lesion,   wherein the first white matter lesion and the second white matter lesion is estimated based on a brain region extracted from a medical image,   wherein the brain region includes white matter from the medical image, wherein the brain region is a region from which a bone region of a brain has been removed.